Skip to main content
Pre-release

This release is a pre-release and may not be stable for production use.

dlin

Crates.io PyPI Ask DeepWiki

dbt model lineage CLI. Parses SQL files directly or reads a compiled manifest.json. No Python required.

Works for developers navigating live SQL files, analysts exploring a shared manifest, AI agents via CLI prompt or MCP server, and CI pipelines.

Column-level lineage (dlin column upstream / dlin column downstream) requires a compiled manifest.json and is usable in practice while remaining Experimental for coverage reasons.

Motivation

When I edited dbt models in VS Code, dbt Power User was my go-to companion for navigating lineage. AI agents have no such companion. I watched them grep through dbt projects to find model dependencies. It works, but they end up calling grep repeatedly and relying on fragile string matching to piece together ref() and source() relationships.

dlin is designed to fill that gap: a CLI tool that lets AI agents understand a dbt project's structure without falling back to grep. It is equally useful for humans, and its stdin/stdout interface makes it easy to combine with jq, git diff, and other CLI tools.

To replace grep, speed and size matter. dlin is a small, self-contained binary with no runtime dependencies. It parses SQL directly, interprets the Jinja that selects dependencies without Python, parallelizes file I/O, and caches manifest and column-lineage analysis.

The key idea behind dlin is that finding the right models fast is what matters most. The hard part for agents is knowing which models to look at in the first place. dlin focuses on making model-level lineage fast and provides column-level lineage for deeper analysis.

Install

Cargo (Rust)

cargo install dlin

pip / uv (Python)

For convenience, dlin is also available as a Python package. The installed binary is native and does not require Python at runtime.

pip install dlin-cli   # or: uv tool install dlin-cli

GitHub Releases

Pre-built binaries for Linux, macOS, and Windows are available on the Releases page. You can also use the installer scripts:

macOS / Linux:

curl --proto '=https' --tlsv1.2 -LsSf https://github.com/eitsupi/dlin/releases/latest/download/dlin-installer.sh | sh

Windows (PowerShell):

powershell -ExecutionPolicy Bypass -c "irm https://github.com/eitsupi/dlin/releases/latest/download/dlin-installer.ps1 | iex"

Quick start

# Full lineage graph
dlin graph -p path/to/dbt/project

# Downstream impact analysis
dlin impact orders

# List models as JSON
dlin list -o json --json-fields unique_id,file_path

# Pipe changed files into lineage
git diff --name-only main | dlin graph -o json

Source modes

dlin supports two source modes for model-level commands.

SQL parse mode (default) is for developers editing a live dbt project. It interprets the Jinja in SQL files and in YAML dependency expressions to find ref() and source() calls without waiting for dbt compile, so model-level lineage quickly identifies the related models. Read the relevant SQL for exact transformation behavior.

Manifest mode (--source manifest) is for analysts or agents analyzing compiled state rather than editing live SQL. A developer compiles once and shares manifest.json; users explore the resolved graph without SQL or Python. Use column-level lineage when a field-level trace is needed; it requires the compiled manifest.

# SQL parse mode: reads SQL files directly (default)
dlin graph orders

# Manifest mode: reads manifest.json only
dlin graph orders --source manifest
dlin summary --source manifest

For model-name inputs, manifest.json is the only file needed in manifest mode. File-path inputs (e.g. models/foo.sql) fall back to standard dbt directory layout when dbt_project.yml is absent, which may not match projects with custom path configuration. (check-manifest always requires a full project.)

Limitations of SQL parse mode

  • var() resolves from vars.yml or dbt_project.yml, then its default (--vars CLI overrides not supported)
  • env_var() uses its default and never reads the environment. A different value at run time can select a different dependency or branch than the one shown
  • Values known only at run time (target, env_var() without a default, run_query() results) do not select dependencies; execute is false, as when dbt parses the project
  • Dependencies selected through adapter.dispatch() or namespaced package macros (e.g. dbt_utils.*) may be missed
  • Generic test IDs are dlin-specific (e.g. test.not_null.orders.order_id) and do not match dbt's naming; use manifest mode when exact test IDs matter

When these limitations matter, use --source manifest.

AI agent integration

CLI approach

Recommended for developers with access to a live dbt project. Works with SQL parse mode (no dbt compile needed) as well as manifest mode.

Install dlin and add the following to your AGENTS.md, CLAUDE.md, or system prompt:

## dbt project structure analysis

Use `dlin` to explore dbt model dependencies.
Do NOT grep/cat/find through SQL files.

```bash
dlin summary                                           # Project overview (start here)
dlin graph <model> -u 2 -d 1 -q                        # Upstream/downstream lineage
dlin impact <model>                                    # Downstream impact with severity
dlin list -o json --json-fields unique_id,sql_content  # Read SQL content
git diff --name-only main | dlin graph -q              # Lineage of changed files
```

For full option reference: `dlin --help`, `dlin graph --help`, etc.

The key line is "Do NOT grep/cat/find through SQL files". Without it, agents default to familiar tools. dlin --help is designed for tool discovery, so the prompt can stay minimal.

MCP server (experimental)

For analysts and agents who work from a distributed manifest.json without access to the full project. Runs in manifest mode only.

dlin mcp exposes a stdio MCP server that AI assistants supporting MCP can connect to directly.

dlin mcp --manifest-path path/to/manifest.json

Available MCP tools: project summary, model search, lineage, impact analysis, and column-level lineage.

The SQL dialect for column lineage is detected from the manifest's adapter_type; --dialect overrides it. Requires a compiled manifest.json (dbt compile).

Features

  • SQL parse mode: single binary, no Python, no manifest.json needed for model-level lineage
  • Manifest mode: works from manifest.json alone; useful for analysts or agents without a full project checkout
  • MCP server (experimental): dlin mcp serves lineage data via stdio MCP for direct AI assistant integration
  • Recursive upstream / downstream: -u N / -d N to control traversal depth
  • Impact analysis with severity: dlin impact scores downstream nodes and flags exposure reachability
  • Composable: stdin accepts model names or file paths; pipe with jq, dlin list, git diff, etc.
  • Agent-friendly: --error-format json emits structured {"level","what","why","hint"} on stderr; --help is designed for tool discovery
  • Column-level lineage (experimental): traces columns across models with transformation classification, reports partial results, where tracing stopped and the assumptions it relied on; requires manifest.json

Mermaid diagrams

dlin outputs Mermaid flowcharts that render natively on GitHub, GitLab, Notion, and other Markdown environments.

Simplified graphs with --collapse

Automatically remove intermediate nodes to see just the endpoints (nodes with no predecessors or no successors); everything in between becomes transitive "(via N)" edges:

# Collapse intermediate models; only endpoints remain
dlin graph --collapse -o mermaid

# Focal mode: keep only sources, exposures, and specified focus models
# (ignores BFS window pseudo-endpoints; works best with -u/-d limits)
dlin graph orders --collapse=focal -u 3 -o mermaid
flowchart LR
    exposure_weekly_report>"weekly_report"]
    model_combined_orders["combined_orders"]
    model_order_summary["order_summary"]
    source_raw_customers(["raw.customers"])
    source_raw_orders(["raw.orders"])
    source_raw_payments(["raw.payments"])

    source_raw_customers ==>|"exposure (via 2)"| exposure_weekly_report
    source_raw_orders ==>|"exposure (via 3)"| exposure_weekly_report
    source_raw_orders -.->|"source (via 1)"| model_combined_orders
    source_raw_orders -.->|"source (via 1)"| model_order_summary
    source_raw_payments ==>|"exposure (via 3)"| exposure_weekly_report
    source_raw_payments -.->|"source (via 1)"| model_order_summary

    classDef model fill:#4A90D9,stroke:#333,color:#fff
    classDef source fill:#27AE60,stroke:#333,color:#fff
    classDef exposure fill:#E74C3C,stroke:#333,color:#fff
    class exposure_weekly_report exposure
    class model_combined_orders model
    class model_order_summary model
    class source_raw_customers source
    class source_raw_orders source
    class source_raw_payments source

Positional focus models are always preserved during collapse, so dlin graph orders --collapse keeps orders even if it would otherwise be intermediate.

Pipe to build focused diagrams

Combine dlin list, jq, and dlin graph to extract exactly the nodes you want:

# Staging models → 1 hop downstream, models only, grouped by directory
dlin list -s 'path:models/staging' -o json | jq -r '.[].label' |
  dlin graph -d 1 --node-type model --group-by directory -o mermaid
flowchart LR
    subgraph models_marts["models/marts"]
        model_combined_orders["combined_orders"]
        model_customers["customers"]
        model_order_summary["order_summary"]
        model_orders["orders"]
    end
    subgraph models_staging["models/staging"]
        model_stg_customers["stg_customers"]
        model_stg_online_orders["stg_online_orders"]
        model_stg_orders["stg_orders"]
        model_stg_payments["stg_payments"]
        model_stg_retail_orders["stg_retail_orders"]
    end

    model_orders -->|ref| model_customers
    model_stg_customers -->|ref| model_customers
    model_stg_online_orders -->|ref| model_combined_orders
    model_stg_orders -->|ref| model_order_summary
    model_stg_orders -->|ref| model_orders
    model_stg_payments -->|ref| model_order_summary
    model_stg_payments -->|ref| model_orders
    model_stg_retail_orders -->|ref| model_combined_orders

    classDef model fill:#4A90D9,stroke:#333,color:#fff
    class model_combined_orders model
    class model_customers model
    class model_order_summary model
    class model_orders model
    class model_stg_customers model
    class model_stg_online_orders model
    class model_stg_orders model
    class model_stg_payments model
    class model_stg_retail_orders model

Column names in nodes with --show-columns

Add --show-columns to include column names inside Mermaid node labels, useful for understanding what each model produces at a glance:

dlin graph orders -u 1 -d 0 --show-columns --node-type model,source -o mermaid
flowchart LR
    model_orders["orders<br/>---<br/>order_id, customer_id, order_date, status, total_amount, payment_method"]
    model_stg_orders["stg_orders<br/>---<br/>order_id, customer_id, order_date, status"]
    model_stg_payments["stg_payments<br/>---<br/>payment_id, order_id, amount, payment_method"]

    model_stg_orders -->|ref| model_orders
    model_stg_payments -->|ref| model_orders

    classDef model fill:#4A90D9,stroke:#333,color:#fff
    class model_orders model
    class model_stg_orders model
    class model_stg_payments model

Combines well with --collapse to show rich detail on fewer endpoint nodes.

Other graph options

dlin graph orders -u 2 -d 1                            # focus on specific model
dlin graph -o mermaid --collapse --show-columns        # columns in collapsed nodes
dlin graph orders --collapse=focal -u 3 -o mermaid    # focal: sources + exposures + orders
dlin graph -o mermaid --group-by directory             # group by directory
dlin graph -o mermaid --direction tb                   # top-to-bottom layout
dlin graph --node-type source,exposure                 # filter by node type
dlin graph -o dot | dot -Tsvg > out.svg                # Graphviz rendering

Output formats: ASCII (default), JSON, Mermaid, Graphviz DOT, Plain, SVG, HTML.

Column-level lineage (Experimental)

dlin column upstream and dlin column downstream trace columns across models. They require a compiled manifest.json, so run dbt compile first. Regression fixtures and real dbt project validation show that column-level lineage is usable in practice. Experimental reflects the Rust stack's shorter production history and remaining cross-dialect and complex-SQL coverage gaps, not a prototype-only status.

Output is JSON by default; -o plain is easier to read, and -o mermaid / -o dot draw diagrams.

# Where does each output column of orders come from?
dlin column upstream orders -o plain

# What downstream columns are affected if stg_orders.order_id changes?
dlin column downstream stg_orders --column order_id

# Mermaid flowchart
dlin column upstream customers -o mermaid
dlin column downstream stg_orders --column order_id -o mermaid

# Specific columns only
dlin column upstream orders --column order_id --column status

# Verify manifest freshness before querying
dlin check-manifest && dlin column upstream orders

Column upstream

Traces each output column of a model back to its source columns, following references across intermediate models. Output columns are listed in SQL order; duplicate names and identifier quoting are kept.

dlin column upstream customers --column order_count --column lifetime_value -o plain
customers (2 columns listed: 2 complete, 0 partial)
  order_count     →  orders.id → stg_orders.order_id → orders.order_id (aggregation)
                       boundary: external relation at stg_orders.order_id for customers.order_count: raw.orders (source, policy: snapshot unsupported; seed external)
  lifetime_value  →  payments.amount → stg_payments.amount → orders.total_amount (aggregation)
                       boundary: external relation at stg_payments.amount for customers.lifetime_value: raw.payments (source, policy: snapshot unsupported; seed external)
  assumes yaml:source.clp.raw.orders (YAML columns taken as exhaustive): order_count
  assumes yaml:source.clp.raw.payments (YAML columns taken as exhaustive): lifetime_value

Each column's lineage is complete, or partial when some contributions could not be resolved. A partial column keeps its known sources and lists the unresolved parts with a reason (for example a column hidden behind an unexpanded SELECT *, an ambiguous name, or unsupported syntax).

A boundary line says where tracing stopped and why. Reaching a source, seed or other external relation, or a constant, is a legitimate end. A model that failed analysis, a missing definition, a cycle, an ambiguous model name or a snapshot is an incomplete stop, and the column stays partial.

An assumes line names a schema the result relies on without proof. By default, a column list declared in YAML is taken as the full list of columns, and the line shows which conclusions depend on that.

Transformation types: direct, aggregation, expression, cast, conditional, unknown.

Column downstream

Finds the downstream models and columns whose values depend on a column.

dlin column downstream stg_payments --column amount -o mermaid
flowchart LR
  subgraph sg0["customers"]
    n0_0["lifetime_value"]
  end
  subgraph sg1["order_enriched"]
    n1_0["amount"]
  end
  subgraph sg2["orders"]
    n2_0["total_amount"]
  end
  subgraph sg3["stg_payments"]
    n3_0["amount"]
  end

  n3_0 -->|"value direct"|n2_0
  n2_0 -->|"value aggregation"|n0_0
  n3_0 -->|"value direct"|n1_0

stg_payments.amount flows directly into orders.total_amount and order_enriched.amount, and orders.total_amount is aggregated into customers.lifetime_value.

Results are split into confirmed impacts, possible impacts reached through a finite set of candidates (such as an ambiguous name), and an exploration status that says whether the search was complete. No impact is concluded only when nothing is confirmed or possible and the exploration is complete. Only value dependencies are reported; columns used only in filter predicates and join conditions (row dependencies) are not collected yet.

Performance benchmark

Correctness-gated means from the synthetic scalability run (3 runs, 1 warmup):

Workload dlin 0.2.4 Parrant 0.17.2 dbt-meta 0.3.8
Volume, 100,150 columns, single upstream 88.9 ms 37.4 s build >120 s timeout; query not run
Deep 64, single upstream 115 ms 1.42 s build 4.86 s; query 411 ms
Fan-out 128, downstream 114 ms 1.78 s build 6.47 s; query 363 ms

These synthetic results use artifacts derived from the real dbt fixture. Native Rust avoids Python startup and dlin showed lower latency here; dlin uses --no-cache, Parrant includes parsing, and dbt-meta query excludes build. They are not a general ranking; see the full results and reproduction.

Known limitations

  • Requires dbt compile: no SQL parse mode fallback; manifest with compiled SQL is always needed
  • Value dependencies only: row dependencies (filter predicates, join conditions) are not collected
  • Partial results: SELECT * over relations without known columns, unsupported syntax and ambiguous names leave lineage partial. In JSON, check output.columns[].lineage.status, output.columns[].lineage.unresolved[], output.shape and boundaries[]; errors[] holds diagnostics such as parse failures
  • YAML column lists are assumed complete: when a YAML column list is used as a schema, it is treated as exhaustive; results that rely on this cite it in evidence[] and show an assumes note in plain, Mermaid and DOT output
  • Dialect: auto-detected from the manifest's adapter_type; pass --dialect to override
  • Cache: results are cached in .dlin_cache/ (--no-cache and --refresh-cache control it)

Key subcommands

list

dlin list                                                   # all models and sources
dlin list orders -o json --json-fields unique_id,file_path  # specific model as JSON
dlin list --node-type source                                # sources only

impact

$ dlin impact orders
Impact Analysis: orders
==================================================
Overall Severity: CRITICAL

Summary:
  Affected models:    1
  Affected tests:     1
  Affected exposures: 1

Impacted Nodes:
  [critical] weekly_report (exposure, distance: 1)
  [high    ] customers (model, distance: 1) [models/marts/customers.sql]
  [low     ] assert_orders_positive_amount (test, distance: 1)

Filtering

dlin graph -s tag:finance,path:marts  # selector expressions (union)
dlin graph --node-type model,source   # filter by node type

Credits

Hard fork of dbt-lineage-viewer by Simon Muller (MIT license). The original focused on TUI-based exploration; dlin removes the TUI and targets non-interactive use: scripting, CI, and AI agents.

License

MIT

Metadata

Release files for dlin-cli 0.3.0a1

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for dlin-cli 0.3.0a1
File Size Uploaded Provenance Actions
dlin_cli-0.3.0a1.tar.gz 758.0 kB Without provenance Details for dlin_cli-0.3.0a1.tar.gz

Built distributions (wheels)

Table of built distributions (wheels) for dlin-cli 0.3.0a1
File Actions
dlin_cli-0.3.0a1-py3-none-win_amd64.whl Python 3 none Windows x86-64 Details for dlin_cli-0.3.0a1-py3-none-win_amd64.whl
dlin_cli-0.3.0a1-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl Python 3 none Linux glibc 2.17+ x86-64 Details for dlin_cli-0.3.0a1-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl
dlin_cli-0.3.0a1-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl Python 3 none Linux glibc 2.17+ ARM64 Details for dlin_cli-0.3.0a1-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl
dlin_cli-0.3.0a1-py3-none-macosx_11_0_arm64.whl Python 3 none macOS 11.0+ ARM64 Details for dlin_cli-0.3.0a1-py3-none-macosx_11_0_arm64.whl
dlin_cli-0.3.0a1-py3-none-macosx_10_12_x86_64.whl Python 3 none macOS 10.12+ x86-64 Details for dlin_cli-0.3.0a1-py3-none-macosx_10_12_x86_64.whl

Total release size: 41.3 MB

Release files / dlin_cli-0.3.0a1.tar.gz

Download URL dlin_cli-0.3.0a1.tar.gz
Size 758.0 kB
Tags Source
SHA-256 checksum
How to use checksums
382ec6f7f26ed08ccd55a87af95446345477e038077c4ec72ccb5edff65c9674
BLAKE2b-256 checksum
How to use checksums
30e462b71cd446b89eed995963e50bb870e9951521037954b27dc86d2d063b21
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via uv/0.13.9 {"installer":{"name":"uv","version":"0.13.9","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"26.04","id":"resolute","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}

Release files / dlin_cli-0.3.0a1-py3-none-win_amd64.whl

Download URL dlin_cli-0.3.0a1-py3-none-win_amd64.whl
Size 8.4 MB
Tags Python 3 Windows x86-64
SHA-256 checksum
How to use checksums
92d947f1425e35c08bfe0ead83da71237eb6c04a922ed7de7a10c4d7790bf9da
BLAKE2b-256 checksum
How to use checksums
5f61cf97fb2c0768fb72dd83df78cf43209081fcb1712a3d57ce5395ccd96601
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via uv/0.13.9 {"installer":{"name":"uv","version":"0.13.9","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"26.04","id":"resolute","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}

Release files / dlin_cli-0.3.0a1-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl

Download URL dlin_cli-0.3.0a1-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl
Size 8.4 MB
Tags Linux glibc 2.17+ x86-64 Python 3
SHA-256 checksum
How to use checksums
cd5afa337d9fc0b03387d9e38439b9483cd21225dc2b5af0b75601fe6a57b627
BLAKE2b-256 checksum
How to use checksums
361c0ba8edaeeaf2cf4a53d524372920a02b8a376673b2c7f7e6e0f4f5ccbf0f
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via uv/0.13.9 {"installer":{"name":"uv","version":"0.13.9","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"26.04","id":"resolute","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}

Release files / dlin_cli-0.3.0a1-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl

Download URL dlin_cli-0.3.0a1-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl
Size 7.9 MB
Tags Linux glibc 2.17+ ARM64 Python 3
SHA-256 checksum
How to use checksums
e004ad8da01d4144a817ae6d442abc2c09c9a03c392a96394632aa35ac14c557
BLAKE2b-256 checksum
How to use checksums
b68cb29af487e0d9063bba14c61c1f8aa7a115410fff073538fb52cf975a270b
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via uv/0.13.9 {"installer":{"name":"uv","version":"0.13.9","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"26.04","id":"resolute","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}

Release files / dlin_cli-0.3.0a1-py3-none-macosx_11_0_arm64.whl

Download URL dlin_cli-0.3.0a1-py3-none-macosx_11_0_arm64.whl
Size 7.7 MB
Tags Python 3 macOS 11.0+ ARM64
SHA-256 checksum
How to use checksums
9f64f6ffe625683fb99c6beecd6573257a27ca15c5a3b9e45c963dbed83e69a0
BLAKE2b-256 checksum
How to use checksums
d25616944d7532149a12bbd06c3b6c12c94d91e3f41c556dc9163a090ed1dbc3
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via uv/0.13.9 {"installer":{"name":"uv","version":"0.13.9","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"26.04","id":"resolute","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}

Release files / dlin_cli-0.3.0a1-py3-none-macosx_10_12_x86_64.whl

Download URL dlin_cli-0.3.0a1-py3-none-macosx_10_12_x86_64.whl
Size 8.1 MB
Tags Python 3 macOS 10.12+ x86-64
SHA-256 checksum
How to use checksums
6c15fe8bf15dbeb906a770d4726894cf025cbe1d2c3ebf7a08556266ad97352b
BLAKE2b-256 checksum
How to use checksums
cad339e8fefa942061b21cebc45df4c2c34f3e20665ac947a80fc0e315bb4f8f
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via uv/0.13.9 {"installer":{"name":"uv","version":"0.13.9","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"26.04","id":"resolute","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}

Security for dlin-cli 0.3.0a1

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page